Blind Channel Estimation Algorithm For MIMO Systems And The Effects Of Estimation And Detection On Channel Capacity

Multi-input multi-output (MIMO) systems have the theoretical channel capacity increasing linearly with the number of antennas, however, when some realistic conditions are considered, the potential capacity gain can hardly be obtained. Among these realistic factors, two points closely related to the receiver are considered in this thesis. That is unideal receiving algorithm and unideal channel estimation. First, to decrease receiver implementation complexity, some unideal detection algorithms usually are adopted in the communication system, thus, causing capacity loss. Second, the theoretical capacity of MIMO channel is derived at the assumption that the channel state information is known at the receiver. Practically, the channel estimation overhead and estimation error will both inevitably decrease the available channel capacity. Based on this analysis, this thesis works on the following two main subjects. In the first part of the thesis we assumes that the MIMO systems resort to training signals for channel estimation and adopt parallel MMSE/ZF detection technique in the receiver. Then taking into consideration the channel estimation overhead, estimation error and capacity loss caused by unideal receive algorithm, we derived the formula for available channel capacity. Based on the analysis of the capacity formula, the optimal training length for MIMO systems with unideal receive algorithm is derived under the assumption that the training signals have equal power as that of the data signals. Then further researches are carried out on the capacity of MIMO systems with optimal training length and parallel MMSE/ZF detection. By analyzing outage probability and capacity distribution, we propose simple antenna selection scheme which decreases outage probability greatly without increasing RF devices. The second part of the thesis focus on MIMO systems with blind channel estimation. Since OFDM technique can greatly reduce receiver complexity, MIMO-OFDM systems are mainly considered in this part. Based on the analysis on correlation of subcarrier channels and correlation of subcarrier signals, we propose two blind channel estimation algorithm structures for MIMO-OFDM systems. By utilizing the sub-channel correlation property induced by multi-path, the first blind estimation structure, we call it structure I, greatly reduced computation complexity and the number of samples required for estimation. The channel parameter can be identified with only one or two OFDM symbols and this makes it suit for fast varying channels. The second blind estimation structure, called structure II, is based on precoding of subcarriers. By using the correlation property of precoded subcarrier signals, we only need to estimate one subcarrier channel, and channels of other subcarriers can be recovered via simple cross-correlation operations. So, structure II has the great advantage of low complexity. Principles of each blind estimation structure are presented in this thesis and the algorithm performance is studied by both theoretical analysis and computer simulation. Then based on capacity analysis of MIMO systems with proposed blind estimation algorithms, we give suggestions on estimation algorithms for different channel environments.